v0.0.381

Release Notes

Version 0.0.380 (May 18, 2026)

• web fetch tool
• custom system prompt for each platform
• add timestamps for completions

Version 0.0.380 (May 17, 2026)

• Tableau performance and reliability improvements

Version 0.0.379 (May 16, 2026)

• fix background completion API
• security patches and fixes

Version 0.0.378 (May 13, 2026)

• Per-member admin-managed note (per-org) injected into the planner prompt as <user_profile> context
• Bulk import members from Excel/CSV with dry-run preview; idempotent — never touches roles or group memberships
• Local password sign-in now works for admins as a break-glass when auth.mode = sso_only
• Cleaner sign-up disabled error message

Version 0.0.377 (May 13, 2026)

• Allow SMTP without credentials (use_credentials: false) for anonymous/open relays

Version 0.0.376 (May 11, 2026)

• Fix connection-indexing crashes ("attached to a different loop" / "unknown protocol state 3") on long Postgres-backed indexing runs by giving the background runner its own NullPool engine

Version 0.0.375 (May 10, 2026)

• Fix MSSQL "0 tables" on case-sensitive / binary collations (e.g. Hebrew_BIN)
• Surface MSSQL schema introspection errors instead of silently returning empty

Version 0.0.374 (May 10, 2026)

• Enrich instructions mgmt and diff
• Fix filter bug in widget preview

Version 0.0.373 (May 8, 2026)

• Query timeout settings
• remove answer tool

Version 0.0.372 (May 7, 2026)

• Fix clarify tool not verbose enough

Version 0.0.371 (May 7, 2026)

• Agent db writes - performance/reliability
• better signal in create data tool
• instructions ui fixes

Version 0.0.370 (May 6, 2026)

• Performance/reliability improvements

Version 0.0.369 (May 3, 2026)

• add usage / quota limits policies organization wide

Version 0.0.368 (May 2, 2026)

• add locale for additional languages
• improve UI for agent mgmt and data soures
• allow upload files (csv/xls/pdf) to agents

Version 0.0.367 (May 1, 2026)

• add a new reindexing connection button
• enable mcp tools by default in org settings
• strengthen clarify tool

Version 0.0.366 (April 27, 2026)

• 70% speed improvements
• better caching for tokens

Version 0.0.365 (April 26, 2026)

• Performance improvements
• Change to a faster token counter approach
• Planner v3 (native Anthropic tool_use) is now the default; set BOW_PLANNER=v2 to fall back to the legacy JSON-envelope planner
• Anthropic prompt caching on planner system prompt + tool catalog; cached_tokens instrumentation for OpenAI/Azure
• Async DB writes for finish_tool_execution + upsert_block_for_tool (next planner call no longer blocks on the prior turn's persistence)
• Measured impact (3/3 trial pass rate, identical plans/SQL): per-trial cost -69% on both Haiku 4.5 and Sonnet 4.6; wall-clock -29% on Sonnet, -5% on Haiku; input tokens -73%

Version 0.0.364 (April 25, 2026)

• feat: evals tools for training mode
• loading mode to when adding new connectins with a large amount of objects
• auto draft new evals when (admin) user thumbs up
• fix bug when submitting a new prompt when completion ends but in agent knowledge harness mode
• added native support to GPT-5.5

Version 0.0.363 (April 22, 2026)

• Improve prompting for Azure default guardrails
• Put oauth in the admin settings
• Improve infer widget visualizations to include filter and agg

Version 0.0.362 (April 20, 2026)

• PBI on-prem server improvements

Version 0.0.361 (April 20, 2026)

• Remove nuxt from prod deployment and serve static files via FastAPI
• feat: add Power BI reporting server (on-prem)
• feat: add Oracle BI integration

Version 0.0.360 (April 19, 2026)

• Fix QVD type parsing
• Improving qvd -> duckdb reliability and performance

Version 0.0.359 (April 19, 2026)

• Enhance Sybase client for better code/timout/error handling
• Add instruction button in Agent panel
• Improve Dockerfile

Version 0.0.358 (April 18, 2026)

• SSO + OBO for data connections: OIDC login now extracts email from the id_token, syncs groups, and propagates user identity through to the warehouse
• Entra ID native support for the On-Behalf-Of flow, including offline_access and hardened OAuth connection handling
• Permission overlay revokes stale rows when a user loses upstream access; data sources returning 403 are skipped instead of failing the run
• SIEM integration with end-to-end test coverage
• Dashboards and Scheduled Tasks promoted to first-class items in the main navigation
• Per-domain signup controls for opening up self-serve access
• New Excel-specific tools for spreadsheet artifacts
• exportCSV() available as a sandbox global so artifacts can produce CSV downloads
• Improved dashboard-generation system prompt for more reliable multi-widget layouts
• Evals harness (dogfooding): YAML suites under tests/evals, pytest runner, LLM matrix from LLM_MODEL_DETAILS, JudgeRule with execution metadata (tokens, iterations, per-tool durations), tag-based filtering, multi-turn support, SSE streaming, and per-turn completions/reasoning in failure reports

Version 0.0.356 (April 13, 2026)

• Share dashboards / conversations with specific users or globally

Version 0.0.356 (April 11, 2026)

• BOW for Excel - you can now have BOW inside your excel!
• PowerBI enhancements

Version 0.0.355 (April 10, 2026)

• Show instruction usage and attribution per turn
• New sidebar in report page to show summary, dashboard and current agent
• New knowledge harness for agentic instruction suggestions
• Faster instructions management
• UI improvements across report and dashboard views
• RBAC: groups, roles, policies, per-data-source permissions, and connection/MCP tools authorization
• LDAP integration for enterprise authentication
• WhatsApp Cloud API integration
• Spider text-to-SQL benchmark eval driver
• Fix: make SMTP password optional in settings
• Added support for a .bowignore file when integrating a git account

Version 0.0.354 (April 5, 2026)

• New Scheduled Tasks: set up recurring or scheduled tasks within reports
• New "Add to Dashboard" button to instantly add widgets to an artifact
• New "Polish" action for quick dashboard refinements
• Show recent queries and artifact shortcuts above the prompt box
• Improved dashboard generation speed and performance
• Improved agent filtering by prioritizing master tables for more reliable results
• Added sandbox support for better agentic code development
• Display abort status during tool execution

Version 0.0.353 (March 30, 2026)

• feat: new a2a integration for timbr
• increase timeout in agent harness

Version 0.0.351 (March 29, 2026)

• WAL mode for SQLite deployments and timeout settings for PostgreSQL
• Performance improvement for the main completion flow
• Add timing metrics across code gen / execution for agent execution traces

Version 0.0.350 (March 29, 2026)

• Add ability to integrate custom MCPs
• Add NetSuite native integration

Version 0.0.349 (March 28, 2026)

• Performance improvements
• additional logging

Version 0.0.348 (March 26, 2026)

• Improve Sybase integration and SQL Anywhere to use tds config

Version 0.0.347 (March 25, 2026)

• Improve context compaction to include inspect_data and set a budget of 200k (overriden by known models if exist)
• Add agent indicator/icon to agent trace
• Add download as png button for charts
• Add more filters to reports page and advanced search

Version 0.0.346 (March 24, 2026)

• Fix bug that images are sent in future completions
• Allow support for secret/access key in Bedrock LLM service

Version 0.0.345 (March 24, 2026)

• Make test_connection and other data client utils async calls

Version 0.0.344 (March 23, 2026)

• Fix artifact sandbox: download React development builds in vendor script
• Remove CDN fallbacks for airgapped deployments — missing vendored libs now fail loudly

Version 0.0.343 (March 22, 2026)

• Set headers/handling for streaming in HTTP calls from front-end
• Improve context mgmt budgeting
• Fork previous created reports

Version 0.0.342 (March 22, 2026)

• Fix context bloat when designing dashboards
• Add full SCIM support
• Enhanced audit trail with more activities
• Expose OpenAPI swagger docs
• Improve animation and frontend look and feel when streaming messages
• Send PDF attachment when publishing a dashboard
• Add read_query tool
• Improve dashboard generation and editing
• BOW for Excel initial set up
• New: GPT-5.4 and GPT-5.4-mini native integration

Version 0.0.341 (March 18, 2026)

• add opentelemetry

Version 0.0.340 (March 18, 2026)

• create/edit artifact tool improvements

Version 0.0.339 (March 17, 2026)

• Sybase connector to support owner schema
• Keep alive for long running MCP queries

Version 0.0.338 (March 16, 2026)

• minor fixes and changes

Version 0.0.337 (March 15, 2026)

• added support for MSSQL 2008 (ODBC 17)
• improve artifact generation (speed and reliability)
• added support for Sisense BI

Version 0.0.336 (March 12, 2026)

• feat: notification service for sending emails — supports dashboard sharing, conversation sharing, and scheduled report delivery with optional PDF attachment

Version 0.0.335 (March 9, 2026)

• fix: improve timbr semantic layer integration
• fix: llm usage chart to show both input and output

Version 0.0.334 (March 8, 2026)

• feat: add support for snowflake semantic views
• fix: improve mssql integration to support schema
• fix: mcp improvements
• Add support for databricks multi-catalog discovery

Version 0.0.332 (March 7, 2026)

• Improved MCP-Apps stability and compatibility with Claude
• Enhanced Databricks SQL connector reliability
• Increased OAuth token storage limits
• Added logging to LLM integrations
• Fix connectivity issues via MCP servers

Version 0.0.330 (March 5, 2026)

• Pre-cache tiktoken encodings in Docker build for airgapped environments
• Added more logging

Version 0.0.328 (March 5, 2026)

• fix: when gpt-5 is in model_id string, don't add temprature

Version 0.0.327 (March 4, 2026)

• Allow skip verify_ssl for custom LLM endpoints
• Intrdouce native Bedrock integration, with IAM/API Key auth methods
• Support MCP-Apps! Now using the MCP in MCP-Apps compatible clients will render visualizations and dashboards
• Introducing Timbr AI beta integration

Version 0.0.325 (March 3, 2026)

• Fix Alembic migration SSL error when using Aurora PostgreSQL with IAM authentication

Version 0.0.324 (March 2, 2026)

• Default SMTP config
• Improve k8s helm to support custom certs when using Aurora DB as backend

Version 0.0.322 (March 1, 2026)

• Support long oauth string columns for Entra
• Allow AWS Aurora PG with IAM as backend DB

Version 0.0.320 (February 24, 2026)

• Improve table lookup
• Improve OAuth MCP integration

Version 0.0.320 (February 22, 2026)

• Support deployment in airgapped systems
• Improve PowerBI integration
• Improve Thumbnail generatio for Artifacts

Version 0.0.319 (February 22, 2026)

• Fixed edit connection "Test Connection" to validate new credentials instead of using saved ones
• Credentials in edit mode are now locked by default with a "Change" button to explicitly unlock
• Renamed "Domains" to "Data Agents" in connection detail modal

Version 0.0.318 (February 22, 2026)

• Added Sybase SQL Anywhere data source connector (enterprise license required)
• Uses FreeTDS ODBC driver for TDS protocol connectivity on port 2638

Version 0.0.316 (February 21, 2026)

• Added filters for low score agent executions in monitoring/diagnosis
• Enhanced file upload and completion context handling, and special support for images
• Pass images and screenshots to create_artifact tool

Version 0.0.315 (February 19, 2026)

• Improved organization logo upload
• Power BI: one table per internal table, relationship support, cleaner SharePoint names

Version 0.0.314 (February 18, 2026)

• Added Microsoft Fabric data source integration (Warehouse and Lakehouse SQL endpoints)
• Azure AD Service Principal authentication support for Fabric
• Added read_artifact tool and improved context engineering for designing dashboards

Version 0.0.313 (February 16, 2026)

• Update license env variable and secret configuration in k8s and docker-compose

Version 0.0.312 (February 14, 2026)

• Refactor sidebar to use nav config and proper active states
• Improved slides artifact generation

Version 0.0.311 (February 13, 2026)

• Multi-connection support: data sources can now have multiple connections
• Added PowerBI and Qlik (QVD) data source integrations (Enterprise)
• Configurable step retention per organization (Enterprise)
• Exclude shared conversations and published reports from step cleanup
• Connection icons shown when describing/inspecting tables
• Schema enrichment with metadata and column comments
• Data agents and example agent templates
• Delete connections support
• Artifact thumbnails
• Added filtering for reports by schedule to easily view reports based on their schedule settings
• Added domain filtering for monitoring diagnosis to filter by specific domains
• Added report thumbnail generation and preview cards on home page for quick visual reference
• Added support for Claude Opus 4.6 model

Version 0.0.309 (February 4, 2026)

• Create artifact (dashboard/slides) tool is now available via MCP
• Added support for Databricks SQL
• Add enterprise license management and audit log

Version 0.0.308 (January 31, 2026)

• Instruction @mentions now only show published instructions from the main build
• Referenced instructions are automatically loaded into AI context when a parent instruction mentions them
• Schema index and full schema now display instruction count per table, guiding the planner to use describe_tables for business rules
• Updated MCP get_context tool to expose instruction count per table
• Microsoft Teams Integration: Full bot support for Teams channels and 1:1 chats
  • Send questions via @mention in channels or direct message the bot
  • Thread-based conversations with report reuse across replies
  • User verification flow with Adaptive Cards
  • Markdown tables, count results, and report links rendered natively in Teams
  • JWT signature verification for inbound webhooks
  • Teams setup UI in Settings > Integrations

Version 0.0.307 (January 28, 2026)

• Separated code and queries for better UX
• Added created/approved by metadata for instructions

Version 0.0.306 (January 26, 2026)

• New Interactive Dashboards: Dashboards are now generated as executable React/HTML code, enabling rich interactivity, custom styling, and dynamic visualizations
• Visual Feedback: Upload screenshots or images with your prompts to show the AI exactly what you want—perfect for requesting design tweaks or pointing out issues
• Dashboard validation now includes automatic screenshot capture, allowing the AI to visually verify the output before finalizing
• Added vision model support for OpenAI, Anthropic, and Google Gemini LLM providers

Version 0.0.305 (January 24, 2026)

• Rebuilt Dashboards: Now fully AI-generated as executable code (React/HTML) with iterative refinement based on conversation history
• Fixed @ mention detection in prompt input (no longer triggers inside existing mentions)

Version 0.0.304 (January 22, 2026)

• SQLite data source now available in production (previously dev-only)
• Security updates and dependency patches

Version 0.0.303 (January 22, 2026)

• AI-suggested instructions now show persistent "Published" status with timestamp
• Added checkbox selection when publishing AI suggestions
• Fixed AI builds not being linked to agent executions

Version 0.0.302 (January 20, 2026)

• Rename Catalog to Queries
• Show chart and visualization in query page

Version 0.0.301 (January 20, 2026)

• Support for local DuckDB databases via file:// or absolute path i.e /data/myduck.db
• Set global git repo management

Version 0.0.300 (January 19, 2026)

• Slack Integration Enhancements
  • Thread-based responses: replies now appear in threads instead of separate messages
  • Each thread corresponds to a single report for better conversation continuity
  • Added support for @mentions in channels (in addition to DMs)
  • Visual feedback via emoji reactions: 👀 when processing, ✅ when complete
  • Data source access control: channel mentions query only public data sources, while DMs include private data sources the user has access to

Version 0.0.298 (January 18, 2026)

• Added guardrails around code execution
• Removed code validation flag, as it's now deterministic and built-in

Version 0.0.297 (January 18, 2026)

• Introducing: Training Mode
  • A dedicated mode for documenting and managing your data domain knowledge
  • Explore schemas, inspect data, and create instructions to guide AI behavior
  • New tools: create_instruction and edit_instruction for real-time instruction management
  • Instructions are versioned and tracked in draft builds until finalized
• Improve DuckDB system prormpt
• HBD!

Version 0.0.296 (January 12, 2026)

• Added PostHog integration for analytics
• Fix Dockerfile

Version 0.0.294 (January 12, 2026)

• improve streaming performance
• support heatmap charts
• block sending prompts if no llm or data source/file were set
• improve conversation layout for mobile presentation
• add delete connection

Version 0.0.293 (January 10, 2026)

• Fix tables page not showing all tables when navigating between pages

Version 0.0.292 (January 9, 2026)

• Fix demo data sources not loading in Docker container

Version 0.0.291 (January 6, 2026)

• Improve streaming for final_message
• Fix multi bar chart rendering bug

Version 0.0.290 (January 1, 2026)

• Happy new year!
• Connections and data sources are now decoupled. You can attach multiple data sources to a single connection, each with its own tables, instructions, and evals. This brings much greater flexibility, reliability, and organization to your workspace.
• New: Context Selector – easily control which data sources are currently active throughout the application.
• Added ability to share report conversations with others
• Clarify tool and prompt optimizations

Version 0.0.288 (December 26, 2025)

• UI improvements: eval, build ID
• Added modal to manage test suites
• Added new MCP tools: list, create, and delete instructions

Version 0.0.286 (December 25, 2025)

• Auto suggest instructions if user provided negative feedback to an answer
• Improve auto-detect uvicorn workers

Version 0.0.284 (December 23, 2025)

• Git providers: Now support Personal Access Token (PAT) authentication for seamless integration.
• You can now create pull requests and branches for build (instruction versions) directly from the interface.
• Each build now includes integration tests and eval runs to ensure greater reliability and code quality.
• Simplified instruction status life cycle and integrating to buid statuses
• UI/UX upgrades: Enhanced workflows for adding instructions and reviewing builds, making navigation and use smoother.
• Code clean ups and tests

Version 0.0.282 (December 22, 2025)

• Launched instruction build/versioning system: every instruction update creates a new version, with point-in-time builds (snapshots), approval workflow, diff, and rollback.
• All instructions now tied to builds; is_main build sets active instruction set for org, with full history & audit.
• Added /builds API: get builds, build diffs, rollback, and detailed version/content lineage for every instruction.
• Test/Eval runs can select which build to use.
• Exposed top-k instructions retrieval API.
• Extensive automated E2E test coverage for build/version/rollback/git flows.

Version 0.0.280 (December 19, 2025)

• Context and instructions are now unified
• Instructions now show detailed usage statistics
• New rules for instruction application: always apply, or smart based on relevance/search
• Instructions table redesigned—now with filters, git-sourced instructions, and other enhancements
• Improved create/edit instruction workflow with a refreshed design
• Expanded and updated automated end-to-end tests

Version 0.0.279 (December 17, 2025)

• Added MCP Server for integration with Claude, Cursor, and other MCP clients
• Available tools: create_report, get_context, inspect_data, create_data
• MCP sessions are fully tracked in reports with tool executions and visualizations
• Added per-user API keys for MCP and external integrations

Version 0.0.278 (December 15, 2025)

• Enhancing MongoDB integration to support Atlas/SRV connections
• Add more triggers for autogenerate suggestions
• UI improvements/fixes

Version 0.0.277 (December 14, 2025)

• Frontend tests (playwright) and CI/CD improvements

Version 0.0.274 (December 12, 2025)

• Added support for GPT-5.2 model
• Enhanced the describe entity tool for better usability and accuracy
• Fixed a user authentication bug affecting specific environments

Version 0.0.271 (December 10, 2025)

• Describe entity from catalog - new tool!
• Remove forgot password/etc when SMTP is not available

Version 0.0.270 (December 10, 2025)

• bug fixes, performance and reliability

Version 0.0.269 (December 10, 2025)

• Performance and speed

Version 0.0.268 (December 9, 2025)

• Speed and readme

Version 0.0.266 (December 8, 2025)

• Added a new Inspect Data tool for quickly examining the structure and sample content of a dataset and preview data before generating insights or diagnosing issues
• Docker Compose now bundled for both development and production environments
• Added sample databases to assist onboarding and demos
• Enhanced overall system reliability and robustness

Version 0.0.265 (December 7, 2025)

• Bug fixes

Version 0.0.264 (December 6, 2025)

• Enhanced file management and analysis capabilities (supports xls, csv, and pdf files)
• Improved MariaDB improvements
• Add support for loading up to 60K tables when connecting data sources
• Added automated tests for postgres database

Version 0.0.263 (December 4, 2025)

• System prompt improvements and a new section for analytical standards
• Improvements to custom LLM integration (set default/small default models)
• Data source onboarding improvement

Version 0.0.262 (December 2, 2025)

• Added data source integration to MongoDB
• Added native support for Custom LLM endpoints (openai compatible)
• Added support for Claude Opus 4.5

Version 0.0.261 (December 2, 2025)

• Bias partitions in bigquery

Version 0.0.260 (December 2, 2025)

• Dependencies updates
• Improve instructions list modal

Version 0.0.259 (December 1, 2025)

• Introducing Filters in dashboards
• Performance improvements, page loads, indices, reliability, and more
• Improved resources selector in context page (toggle between chunks/files, index status info, and more)
• UI enhancements

Version 0.0.258 (December 1, 2025)

• Increase anthropic max tokens to 32k
• Impove behavior of reindexing (do not auto-add)

Version 0.0.257 (November 30, 2025)

• Added Azure Data Explorer data source (thanks @licanhua)
• Improved BigQuery system prompt to consider special syntax guidelines when generating code

Version 0.0.256 (November 29, 2025)

• Improved visualization features
• Enhanced dashboard creation workflow
• Suggestions now cover more user actions, such as corrections, querying the same tables, and sharing code
• Expanded instruction categories for system, dashboard, and visualizations
• UI improvements for agent trace, observations, and reduced visualization flicker
• Improved data source onboarding and test connections
• Added integration tests for LLMs and popular data sources

Version 0.0.255 (November 27, 2025)

• Extended user token validity to one week, reducing the need for frequent logins
• Improved evaluation (Evals) features for more robust and insightful testing
• Added support for anonymous MySQL connections

Version 0.0.254 (Noveber 25, 2025)

• Fix azure llm integration
• Improve mysql authentication

Version 0.0.253 (November 24, 2025)

• Gemini 3 Pro Preview added!

Version 0.0.252 (November 22, 2025)

• Implemented tracking of LLM usage and associated costs in the console dashboard
• Enhanced metadata resource handling:
  • Remove objects no longer found during reindexing
  • Newly discovered objects are no longer auto-activated by default
• Introduced SQLite integration (for testing and development), and expanded test coverage for git repositories, metadata resources, and more
• Improved the process for deleting data sources
• Added bulk archive functionality for reports and revamped the main reports index page

Version 0.0.251 (November 20, 2025)

• Data sources deletion

Version 0.0.250 (November 19, 2025)

• Add context estimator when writing prompts

Version 0.0.249 (November 19, 2025)

• Pinot get tables to use user:pass when creating the HTTP request

Version 0.0.248 (November 18, 2025)

• Resolve flickering in the Reasoning section and enhance the reliability of data source deletion and modal overlays
• Improve stability and robustness of table auto-activation and deactivation

Version 0.0.247 (November 17, 2025)

• Instruction labels added for more effective categorization and management
• Instructions can now be auto-enhanced with AI suggestions
• Message display now clearly distinguishes between user and agent responses
• Trace modal correctly navigates to the selected completion ID within the reports page

Version 0.0.246 (November 16, 2025)

• Snowflake keypair auth
• Repair migrations

Version 0.0.245 (November 16, 2025)

• Repair migrations

Version 0.0.244 (November 15, 2025)

• Updating dependencies

Version 0.0.243 (November 15, 2025)

• Fixing a couple of bugs and renaming release notes to CHANGELOG

Version 0.0.242 (November 14, 2025)

• Enhanced markdown parser for better handling of complex formatting and edge cases
• Added support for Dataform projects and introduced SQLX file parsing, enriching contextual metadata for queries and models
• Integrated GPT-5.1 as an available LLM by default
• Improved metadata indexing service with additional guardrails for git repository management and error management
• Upgraded user interface for reports and tables

Version 0.0.241 (November 14, 2025)

• Optimize datbase migrations to include report_type
• Wrap maintenance job with guardrails

Version 0.0.240 (November 13, 2025)

• Introducing Evals! You can now create and run custom sets of tests on demand to assess system performance. Define your own test cases and assertions, such as:
  • User prompts triggering create_data on table1 and table2
  • Validating that specific data columns (e.g., a, b, c) are present
  • Using custom LLM Judge prompts to automatically determine pass/fail outcomes
• Added the ability to adjust the sample k size for schema tables and metadata resources
• Improved the data source pages for a faster, smoother experience, including enhanced loading indicators and improved item removal
• Unused steps are now auto-deleted after 14 days. You can restore them anytime by rerunning the code.

Version 0.0.236 (November 13, 2025)

• Added sorting and filtering capabilities to the table selector
• Reduced logging verbosity in production environments
• Enforced strict limits on context section sizes

Version 0.0.235 (November 12, 2025)

• Added ability to select and deselect items in table and metadata resource selectors
• Enhanced BigQuery integration to allow connections to multiple datasets
• Enforced organization-level uniqueness for data source and LLM provider names
• Allow service json for BigQuery required user auth mode

Version 0.0.233 (November 11, 2025)

• Improved instructions visibility in prompts' context
• Introduced an "Analysis Panel" for admins when creating or approving instructions:
  • Impact Score Estimation: Evaluate how the new instruction relates to existing prompts and user questions
  • Related Instructions: Identify potential redundancy or conflicts with other instructions
  • Related Metadata Resources: Review if the instruction overlaps or conflicts with current enriched context (such as dbt, markdown, etc.)

Version 0.0.232 (November 10, 2025)

• Introduced default small models: you can now designate a default "small" model for back-office operations such as evals, judge tasks, instruction generation, and more
• User feedback (thumbs up/down) is now attributed at the table level

Version 0.0.231 (November 8, 2025)

• Enhanced the UI for agentic retrieval and search for greater clarity and usability
• Refined the agent head prompt to more effectively leverage and guide the use of search tools
• Improved the agent trace user interface for better readability and interaction

Version 0.0.230 (November 6, 2025)

• Introduced a new create_data tool that is more robust, reliable and accurate data generation
• Enhanced code generation for more accurate and robust SQL and Python outputs
• Improved chart visualizations for clearer and more informative data presentation
• Added new data source integration support: Apache Pinot and Oracle DB
• Table browsing now displays detailed statistics, including usage frequency, scoring, and feedback metrics
• Launched the new read_resources tool for intelligent, on-demand searching across all metadata resources
• Added successful executed queries in the same tables for when agent is generating code

Version 0.0.220 (November 4, 2025)

• Added BigQuery support for maximum_bytes_billed for cost guardrails and support for use_query_cache
• Improved main AI loop with additional observations from sub-agent create data (code, errors, etc)
• Improved UI for list of instrusctions modal - pagination, visibility, etc

Version 0.0.219 (November 3, 2025)

• Improved table discovery and retrieval in main agent loop
• Introduced describe_tables tool for better data modeling, with light UI signaling
• Reduced the main agent's context footprint by 5x, significantly faster and leaner
• The create data sub-agent now receives a provided list of tables instead of inferring the data model itself

Version 0.0.218 (November 1, 2025)

• Fixed issue where the data source form was not fully rendered in the onboarding screen
• Fixed issue where Claude outputs a Python code fence before the actual code

Version 0.0.217 (November 1, 2025)

• Basic telemetry (configurable in bow-config)

Version 0.0.215 (October 31, 2025)

• Support multi schema for Postgres client

Version 0.0.214 (October 30, 2025)

• Support multi-db connection for ClickHouse

Version 0.0.213 (October 28, 2025)

• Clickhouse fix
• Better rendering of booleans in connection form

Version 0.0.212 (October 20, 2025)

• Integrate Mentions component and enhance prompt capabilities
• Implement mentions context integration in tools and agents
• Released: Catalog feature for efficient management and discovery of models, metrics, visualizations, and queries. Enables reusable components and enhances AI analyst intelligence
• Fix yarn cache issue in docker image

Version 0.0.206 (October 19, 2025)

• Bug fix reloading tables in schema

Version 0.0.205 (October 17, 2025)

• Added support for multiple schemas in Snowflake
• Added MSSQL driver into Dockerfile

Version 0.0.204 (October 16, 2025)

• Fixed permission issue in Docker when uploading files
• Fixed instructions not showing creator in instruction list

Version 0.0.203 (October 12, 2025)

• Enhanced the chat interaction and conversation flow with the AI agent
  • Improved prompt capabilities by auto setting thinking levels
  • Enhanced message context with processed data and answer metadata for better LLM interactions
• Optimized CI/CD workflows by integrating GitHub Release automation

Version 0.0.202 (October 8th, 2025)

• Added DuckDB support for object store files (aws, gcs, azure)
• Added Claude Sonnet 4.5 support

Version 0.0.200 (September 27, 2025)

• Enhanced data source setup experience for new users
• Redesigned user interface for data source management
• Introduced "require user authentication" option for data sources
• Sample questions for data sources is now customizable
• Added to organizations ability to set judge, autogen instructions and code editing as enabled/disabled
• Added a bunch of AGENTS.md files throughout the repo for faster and better coding

Version 0.0.199 (September 20, 2025)

• Redesigned application onboarding experience
• Implemented automatic instruction suggestions throughout the onboarding process
• Added support to Tableau as a data source
• Some general updates, bug fixes and new tests and sentry removal

Version 0.0.198 (September 17, 2025)

• Adding login with OpenID Connect (Okta, etc)
• Updating Helm to allow oidc params and auth mode (hybrid, local or sso)
• Touch up to signin/signup screens
• Fix docker image to include client for openssh

Version 0.0.197 (September 15, 2025)

• Introduced Tableau data source integration: TDS files can now be imported to enhance contextual information for data sources
• Deprecated AI Rules feature at the data source level, consolidating rule management into the centralized instruction system
• Added support for Google Gemini LLM
• Added verbosity to git integration
• Squashed bugs and improved overall usability

Version 0.0.196 (September 14, 2025)

• Added inline code editor for queries with full execution capabilities: users can now edit query code, preview data results, visualize outputs, and save changes directly within the interface
• Added widget customization controls for labels, titles, and styling
• Rebuilt query/visualization engine for improved scalability
• Improved dashboard layout, reactivness and synchronization to other visualizations
• Enhanced backend architecture and data modeling to support query versioning and multi-visualization relations
• Added ability to test LLM connection before saving as a new provider

Version 0.0.195 (September 10, 2025)

• Introducing Deep Analysis: Users can now change from Chat mode to Deep Analytics for doing a more comprehensive open ended analytics research to identify root cause, anomalies, opportunities, and more!
• New Prompt box for both home/report page, including customizing LLM per prompt
• Roles with console/monitoring access can now view the full agent loop trace inside the report chat

Version 0.0.194 (September 9, 2025)

• Enhanced Dashboards
  • Improved dashboard creation, allowing more control on styles and the new dashboards look amazing!
  • User can now select themes (default, retro, hacker, or research)
• Added the answer question tool, allowing agent to search across schema, resources, and other pieces context to come up with the answer
• Improvements to Slack bot integration
• Enhancements around: cron visibility, excel files, and sharing

Version 0.0.193 (September 6, 2025)

• Introduced automatic instruction suggestion system to enhance AI decision-making and performance. The system generates suggestions triggered by:
  • User clarifications regarding terms, facts, or metrics
  • AI successfully resolving data generation code after encountering multiple failures
• All generated suggestions are stored globally and require administrative review and approval before implementation
• Improved main AI agent planner prompt
• Redesigned and expanded the navigation menu, elevating monitoring and instructions to prominent first-class menu items
• Bug fixes and enhancements

Version 0.0.192

• Fixed file upload functionality within Docker container environment
• Resolved issues with report rerunning capabilities
• Reduced database logging output to only display warnings and errors

Version 0.0.190 (August 31, 2025)

• Launched Agent 2.0, a comprehensive redesign of the backend agentic architecture
  • Implements ReAct methodology with single-tool execution per planning cycle
  • Enhanced tool registry featuring comprehensive tracking and governance capabilities
  • Added clarify tool for detecting user queries with undefined metrics/measures or ambiguous requirements
  • Improved error handling, tool schema validation, and enhanced reliability throughout agent execution
  • Comprehensive tracking system for agent executions, tool usage, and AI decision-making processes
• Released Context Management 1.0, providing robust and reliable context tracking for both warm and cold AI interactions
  • Complete monitoring of context utilization patterns
  • Streamlined interface for context construction and management during agent operations
• Enhanced compatibility with LLMs that generate prefix/postfix formatting symbols such as json/``` markers
• Redesigned streaming architecture with server-sent events (SSE) implementation for real-time user prompt processing
• Enhanced admin interface for monitoring agent execution flows and tracking user request patterns
• Introduced new analytics visualization in console dashboard displaying metrics for data request creation (user-initiated), AI clarification requests, and additional operational insights
• Added automated testing for the system
• As this change was signifcant, old reports (in version prior 0.0.190) will be set as read-only.
• Introduced customizable branding and AI identity features, allowing organizations to upload their own logos, remove Bow attribution, and personalize their AI assistant's identity

Version 0.0.189 (August 25, 2025)

• Enhanced table usage analytics with comprehensive success/failure tracking, performance scoring, and intelligent usage pattern recognition
• Implemented automated TableStats model to capture query performance metrics, execution outcomes, and user satisfaction data in real-time
• Advanced code generation now leverages historical success patterns and proven code snippets, significantly improving accuracy and reliability
• Upgraded AI planner with feedback-driven decision algorithms that incorporate table performance scores and usage data for continuous self-improvement
• Added weighted performance/feedback scoring based on user role (admin vs. rest)
• Added tests covering llm providers, azure backend, and console metrics

Version 0.0.188 (August 23, 2025)

• Enhanced streaming reliability for data models and query results in chat interface
• Strengthened completion termination handling with comprehensive SIGKILL support across all agent lifecycle stages
• Introduced custom base URL configuration for OpenAI provider deployments
• Resolved console metrics and usage data functionality issues
• Corrected admin permissions to allow deletion (not just archival/rejection) of suggested instructions

Version 0.0.186 (August 19, 2025)

• Enhanced instructions functionality with support for referencing dbt models, tables and other metadata resources
• Updated data source section with improved views of dbt and other metadata resources
• Fixed various bugs and enhanced overall usability

Version 0.0.181 (August 10, 2025)

• Added data source visibility controls - admins can now set data sources as public or private within organizations and manage granular access permissions through user memberships
• Improved interface and user experience with differentiated views and controls for administrators versus regular users in the data source management area
• Integrated OpenAI's latest GPT-5 language model into the platform
• Updated Docker image to use Ubuntu base with latest security patches
• Updated Python package dependencies to latest stable versions
• Implemented container vulnerability scanning using Trivy in CI/CD pipeline

Version 0.0.180 (August 6, 2025)

• Enhanced security by updating Dockerfile with latest vulnerability patches
• Integrated Claude 4 Sonnet and Opus language models
• Implemented full support for Vertica database connectivity and querying
• Added capability to incorporate markdown files from git repositories to enhance data sources with contextual information
• Added support for Azure OpenAI and custom model endpoints
• Added support for AWS Redshift database connectivity

Version 0.0.177 (July 30, 2025)

• Added comprehensive admin console with three main sections: Explore, Diagnose, and Instructions management
• Explore: Organization analytics dashboard with real-time metrics, activity charts, performance tracking, table usage analysis, table joins heatmap, failed queries overview, recent instructions, top users, and prompt type analytics
• Diagnose: Advanced troubleshooting interface featuring failed query tracking, negative feedback analysis, instructions effectiveness scoring, detailed trace debugging, and issue categorization with actionable insights
• Instructions: Centralized instruction management system with search and filtering capabilities, add/edit functionality, data source associations, and user permission controls
• Added LLM Judge system for automated quality assessment - scores instruction effectiveness and context relevance on a 1-5 scale, evaluates AI response quality against user intent, and provides detailed reasoning for continuous system improvement

Version 0.0.176 (July 26, 2025)

• Added ability to provide detailed feedback messages when submitting negative feedback on AI completions
• Improved reports main page UI

Version 0.0.175 (July 26, 2025)

• Added ability for users to suggest new instructions and view published instructions
• Added workflow for admins and privileged users to review, approve, or reject suggested instructions
• Enhanced instruction management with data source associations - instructions can now be set globally or scoped to specific data sources
• Added visibility controls allowing admins to hide certain instructions from unprivileged users

Version 0.0.174 (July 23rd, 2025)

• Filters and pagination for reports
• Reports are now invisible for other users when not published

Version 0.0.172 (July 17th, 2025)

• Slack integration! Now admins can integrate their Slack organization account and have users converse with bow via slack. Includes user-level authorization, formatting, charts, and tables
• LookML support for git integration indexing
• Download steps data as CSV is now available in UI
• Added Instructions: add custom rules and instructions for LLM calls

Version 0.0.166 (July 13th, 2025)

• Resolved membership invitation handling for closed deployments with OAuth authentication
• Corrected query count calculation in admin dashboard metrics

Version 0.0.165 (July 7th, 2025)

• Added admin dashboard with usage analytics, query history tracking, and LLM feedback collection
• Implemented secure password recovery workflow with email verification
• Enhanced Kubernetes deployment configuration with expanded Helm chart coverage and options

Version 0.0.164 (April 24th, 2025)

• Refactored dashboard visualization capabilities:
  • Improved chart rendering performance and responsiveness
  • Enhanced data handling for large datasets
  • Added better error handling and validation
  • Streamlined chart configuration options
• Fixed candlestick chart bug where single stock data was not properly displayed when no ticker field was present
• Added "File" top level navigation item. You can now see all files uploaded in the org
• You can now mention files outside of the report
• Support older version of Excel (97-03)

Version 0.0.163 (April 21, 2025)

• Added new charts: area, map, treemap, heatmap, candletick, and more
• Better experience for charts to handle zoom, resize and overall better rendering

Version 0.0.162 (April 16, 2025)

• Added ability to stop AI generation mid-completion with a graceful shutdown option
• Enhanced application startup reliability with automatic database connection retries
• Moved configuration management to server-side, enabling centralized client configuration
• Introduced support for deploying the application on Kubernetes clusters using Helm charts

Version 0.0.161 (April 14, 2025)

• Added support to OpenAI GPT-4.1 model series

Version 0.0.160 (April 12, 2025)

• Enhanced AI reasoning with ReAct framework and advanced planning capabilities
• Added upvote/downvote system for users to provide feedback on AI responses
• Added detailed reasoning explanations for AI responses in both UI and backend
• Improved Completion API to support synchronous jobs and return multiple completions
• Added OpenAPI support for global authentication and organization ID handling
• Enhanced organization settings and key management system
• Added visual source tracing in data modeling interface

Version 0.0.155 (March 30, 2025)

• Added code validation for generated code
• Added safeguards for planner and coder agents
• Enabled code review for user's own code
• Fixed memory bug
• Added reasoning for planner agent
• Added data preview for LLM to achieve ReAct like flow with code generation
• Added organization settings to control AI features (specific agent skills) and additional settings (LLM viewing data, etc)
• Added df summary for tables
• Refactored code execution to be more robust and handle edge cases better

Version 0.0.154 (March 24, 2025)

• Added advanced logging infrastructure
• Added e2e tests infrastructure and created first e2e test for user onboarding
• Improved ci/cd to run tests before building image

Version 0.0.153 (March 22, 2025)

• Added support with dbt (via git repo) models and metrics
• Added context building for dbt models
• Added token usage to plan
• Added x-ray view for completions for admin roles

Version 0.0.152 (March 16, 2025)

• Added AWS Athena integration
• Fixed bug when generating data source items
• Fixed bug when deleting data sources

Version 0.0.151 (February 25, 2025)

• Added Claude 3.7 Sonnet to LLM models
• Added sync provider with latest models

Version 0.0.15 (February 24, 2025)

• Added active toggle to data source tables to hide from context
• Fixed bug when generating data source items
• Add top bar to index page when no LLMs are available

Version 0.0.14 (January 3, 2025)

• Added basic self-hosting support
• Added printing in code gen for better healing
• Improve answering agent and planner agent
• Replaced highcharts with ECharts
• Added intercom
• Various fixes and improvements

Version 0.0.13 (December 26, 2024)

• Added prompt guidelines
• Fixed modify, creation of widgets
• Fix proxy in nuxt/fastapi
• Improved agents: dashboard, data model, chart, and prompt
• Added email validation for signups
• Dockerized the application
• Kubernetesized the application

Version 0.0.12 (December 13, 2024)

• Added functionality to rerun dashboard steps, including cron support with configurable intervals
• Enabled automated LLM-generated summaries, starters, and reports for connected data sources
• Integrated Google Sign-In for seamless user authentication
• Added support for nginx reverse proxy
• Redesigned the home page for improved usability
• Added bow-llm, an abstracted LLM provider to set as the default
• Enhanced error handling with interactive toasts for better feedback
• Improved agent capabilities for code generation with better data source context and refined JSON parsing
• Enabled dynamic modifications to agent plans
• Resolved the "thinking bug"
• Made LLM provider presets uneditable
• Fixed WebSocket functionality in production
• Completed end-to-end tests for completions and data sources

Version 0.0.11 (December 5, 2024)

• Completed integrations for Presto, Salesforce, and Google Analytics
• Added support to CRUD model providers and LLM models
• Added Claude AI model support
• Implemented data source credential security
• Enhanced agent capabilities:
  • Added clarification questions feature
  • Fixed dashboard layout generation
  • Fixed chart parameter rendering
  • Improved data model modifications
• UI Improvements:
  • Fixed report title updates
  • Resolved copy-paste styling issues in prompt box
  • Completed memberships interface
  • Enhanced mention component
• Infrastructure updates:
  • Added configuration file support
  • Removed Excel special routes
  • Cleaned up Nuxt from git repository
  • Fixed default menu data source association
  • Removed unique organization name requirement

Version 0.0.10 (November 28, 2024)

• Edge left menu is now scrollable.
• Fixed logo scaling issue in Edge browser.
• Added schema browser for data sources.
• Enabled manual test connection for data sources.
• Converted data source list in prompts to a dictionary for better position handling.
• Added Markdown support for completions in both agent and UI.
• MySQL, BigQuery, Snowflake, MariaDB, and ClickHouse integrations are complete.
• Initial scaffold for service type data sources
• Fixed _build_schemas_context to run only once during agent initialization.
• Improved data source error messages.
• Only active data sources are now displayed.
• Data sources failing test connection are automatically set to inactive.
• Introduced a service-type architecture for data source handling in code generation.
• Permissions module completed.
• Public dashboard completed.